Skip to content

Comment on Parsing – REBOL Users Guide

Comments

Reminds me of perl6. Why don't more languages support grammars like this?

Fun example: https://github.com/perl6/perl6-examples/blob/master/categori...

Here's that fun example written in Rebol parse:

    digits:     charset "1234567890"
    char:       charset [#"a" - #"z" "+/-*"]  ;; chars allowed in a symbol
    not-quot:   complement charset {"}
    esc-string: [{\"} | not-quot]

    bool:   ["#t" | "#f"]
    float:  [opt "+-" any digits "." some digits]
    number: [opt "+-" some digits]
    symbol: [some char]
    quot:   [#"'" [sexp | atom]]
    string: [{"} any esc-string {"}]

    statement: [sexp | atom | space | newline]
    sexp: ["(" any statement ")"]
    atom: [bool | float | number | quot | symbol | string]

    lisp-rule: [some statement]
NB. I wrote this "off the cuff" so I don't expect it to be flawless! Also there are no capture rules provided in my example (unlike perl6 grammars which provides this automatically).

It's worth noting that this grammar is a bit more restrictive than the Perl 6 equivalent. In particular, "\d" in Perl includes not only 0 through 9, but any Unicode character that's classified as a "digit". The Perl 6 grammar documentation features this in an example of using arbitrary methods in grammars: https://docs.perl6.org/language/grammars.html#Methods_in_Gra...

I think the Perl 6 example should really use [0-9] instead of \d because it's unlikely that unicode numerals are allowed (in the Common Lisp spec).

Anyway neither example fully covers the Common Lisp spec for a number atom (for eg. scientific notation or rationals).

I don't think covering all of the Common Lisp spec is required to be Greenspun-compliant ;)

In that case Rebol is already Greenspun-compliant ;-)

    (print (multiply (add 1 2) 3))
... will print an answer of 9

Usually they do through a PEG library if you search for that. Lua has a nice one and Python a few ok ones. They aren't native to the language like Rebol and Perl6 though.

Yeah, I meant natively.

Oh, yea. P6, Rebol, Red...any lisp with a little work.

And Perl5. Here are two examples of grammar engines built out of perl regex...

* Regexp::Grammars - https://metacpan.org/pod/Regexp::Grammars

* Parse::RecDescent - https://metacpan.org/pod/Parse::RecDescent

Yea, not native which is what the poster ended up meaning, but I hear that one you linked from Damian Conway is great.

No external library is used in these modules just Perl5 native regex engine.

NB. I think thats also same for Parsec (built out of Haskell's parser combinators)

BTW - Damian created both of them, Regexp::Grammars was his replacement for his Parse::RecDescent.

Interesting. I didn't know they had no dependencies. Is it still considered native though since you have to import another module? Rebol & Perl6 don't require that although I'm probably grasping at straws lol. Thanks for pointing out Damian replaced Parse::RecDescent with the other one as I was reading doc on it today at work.

You can write grammars in pure Perl5 regex but it can get a bit fugly :)

Here's a (very) simple example I did in Perl5 - https://news.ycombinator.com/item?id=6895126

And here's a (nicer!) port to Rebol I also did - https://www.reddit.com/r/programming/comments/1smpa1/why_reb...

Thank you!

Rascal MPL is another one that comes to mind.

http://www.rascal-mpl.org/

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.